tablassert 7.3.1__tar.gz → 7.3.3__tar.gz

{tablassert-7.3.1 → tablassert-7.3.3}/AGENTS.md

@@ -35,7 +35,7 @@ src/tablassert/
   lib.py          # Core logic: encodings, data loading, Tcode(Section) class
   models.py       # Pydantic v2 models (TablaBase base class)
   enums.py        # str, Enum subclasses (Tokens, Repositories, Comparisons, etc.)
-  fullmap.py      # NER / entity resolution (DuckDB, 16 shards)
+  fullmap.py      # NER / entity resolution (DuckDB, 12 shards)
   qc.py           # Quality control (ONNX/BioBERT, sentence_transformers)
   nlp.py          # Text normalization (level_one: strip+lowercase, level_two: regex)
   ingests.py      # YAML ingestion: from_yaml(), to_sections(), fastmerge()

{tablassert-7.3.1 → tablassert-7.3.3}/CHANGELOG.md

@@ -2,6 +2,19 @@
 
 All notable changes to this project are documented in this file.
 
+## 7.3.3 - 2026-04-08
+
+### Bug Fixes
+- Changed datassert shard count from 16 to 12 (`SHARDS` constant in `fullmap.py`) to correspond to the updated datassert database layout.
+
+### Documentation
+- Updated all shard count references across documentation and examples to reflect the new 12-shard datassert layout.
+
+## 7.3.2 - 2026-04-03
+
+### Maintenance
+- Updated dependencies. No API changes.
+
 ## 7.3.1 - 2026-04-03
 
 ### Changes

{tablassert-7.3.1 → tablassert-7.3.3}/CONTRIBUTING.md

@@ -20,15 +20,12 @@ cd Tablassert
 uv sync
 ```
 
-### Optional Dependency Groups
+### Optional Extras
 
-Some features require optional dependencies:
+All ML, web, and Excel dependencies are included in the core install. The only optional extra is a runtime-compatible Polars build for CPUs without required instructions:
 
 ```bash
-uv sync --extra ml        # sentence-transformers, onnxruntime, scikit-learn
-uv sync --extra web        # playwright
-uv sync --extra pyexcel    # pyexcel
-uv sync --extra full       # all optional deps
+uv sync --extra rtcompat   # polars[rtcompat]
 ```
 
 ## Development Workflow

{tablassert-7.3.1 → tablassert-7.3.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tablassert
-Version: 7.3.1
+Version: 7.3.3
 Summary: Extract knowledge assertions from tabular data into NCATS Translator-compliant KGX NDJSON — declaratively, with entity resolution and quality control built in.
 Project-URL: Homepage, https://github.com/SkyeAv/Tablassert
 Project-URL: Source, https://github.com/SkyeAv/Tablassert
@@ -42,7 +42,7 @@ Requires-Dist: rapidfuzz>=3.14.3
 Requires-Dist: scikit-learn>=1.8.0
 Requires-Dist: sentence-transformers>=5.3.0
 Requires-Dist: sqlite-utils>=3.39
-Requires-Dist: typer>=0.24.1
+Requires-Dist: typer>=0.21.2
 Requires-Dist: xxhash>=3.6.0
 Provides-Extra: rt
 Requires-Dist: polars[rtcompat]>=1.39.0; extra == 'rt'
@@ -99,7 +99,7 @@ docker run --rm \
 # Build a knowledge graph from a YAML configuration
 $ tablassert build-knowledge-graph graph-config.yaml
 ⠋ Loading table configurations...
-⠋ Resolving entities across 16 DuckDB shards...
+⠋ Resolving entities across 12 DuckDB shards...
 ⠋ Compiling subgraphs...
 ⠋ Deduplicating nodes and edges...
 ✓ Done — wrote nodes.ndjson and edges.ndjson to .storassert/

{tablassert-7.3.1 → tablassert-7.3.3}/README.md

@@ -47,7 +47,7 @@ docker run --rm \
 # Build a knowledge graph from a YAML configuration
 $ tablassert build-knowledge-graph graph-config.yaml
 ⠋ Loading table configurations...
-⠋ Resolving entities across 16 DuckDB shards...
+⠋ Resolving entities across 12 DuckDB shards...
 ⠋ Compiling subgraphs...
 ⠋ Deduplicating nodes and edges...
 ✓ Done — wrote nodes.ndjson and edges.ndjson to .storassert/

{tablassert-7.3.1 → tablassert-7.3.3}/docs/api/fullmap.md

@@ -36,7 +36,7 @@ Column name containing text strings to resolve.
 
 **`conns: list[object]`**
 
-List of 16 DuckDB shard connections to the datassert database.
+List of 12 DuckDB shard connections to the datassert database.
 
 Each shard contains:
 - Synonym mappings (text → CURIE)
@@ -125,11 +125,11 @@ from tablassert.enums import Categories
 import duckdb
 import polars as pl
 
-# Open all 16 shard connections
+# Open all 12 shard connections
 datassert_dir = "/path/to/datassert"
 conns = [
     duckdb.connect(f"{datassert_dir}/data/{i}.duckdb", read_only=True)
-    for i in range(16)
+    for i in range(12)
 ]
 
 # LazyFrame with data to resolve
@@ -167,11 +167,11 @@ from tablassert.fullmap import resolve
 from tablassert.nlp import level_one, level_two
 from tablassert.enums import Categories
 
-# Open all 16 shard connections
+# Open all 12 shard connections
 datassert_dir = "/path/to/datassert"
 conns = [
     duckdb.connect(f"{datassert_dir}/data/{i}.duckdb", read_only=True)
-    for i in range(16)
+    for i in range(12)
 ]
 
 # Map a list of gene symbols to CURIEs

{tablassert-7.3.1 → tablassert-7.3.3}/docs/api/lib.md

@@ -2,7 +2,7 @@
 
 The `lib` module exposes `resolve_many()`, a high-level convenience function for resolving an iterable of entity strings to CURIEs without requiring manual LazyFrame construction, NLP preprocessing, or DuckDB shard management.
 
-It wraps the lower-level [`resolve()`](fullmap.md) pipeline — applying `level_one` and `level_two` normalization, opening all 16 DuckDB shard connections, executing entity resolution, and returning results as a plain Python dictionary.
+It wraps the lower-level [`resolve()`](fullmap.md) pipeline — applying `level_one` and `level_two` normalization, opening all 12 DuckDB shard connections, executing entity resolution, and returning results as a plain Python dictionary.
 
 ## resolve_many()
 
@@ -38,7 +38,7 @@ Examples: `["TP53", "BRCA1", "EGFR"]`, `("aspirin", "ibuprofen")`, or a generato
 
 **`datassert: Path`**
 
-Filesystem path to the root of the datassert database directory. The function expects a `data/` subdirectory containing 16 DuckDB shard files (`0.duckdb` through `15.duckdb`).
+Filesystem path to the root of the datassert database directory. The function expects a `data/` subdirectory containing 12 DuckDB shard files (`0.duckdb` through `11.duckdb`).
 
 Each shard contains:
 - Synonym mappings (text → CURIE)
@@ -98,7 +98,7 @@ Each dictionary contains the following keys (where `{col}` is the value of the `
 
 2. **NLP normalization** — Applies `level_one()` (whitespace stripping + lowercasing) and `level_two()` (non-word character removal via `\W+`) to produce the two normalized columns required by `resolve()`.
 
-3. **DuckDB connection management** — Opens all 16 shard connections inside a `contextlib.ExitStack`, ensuring every connection is properly closed when resolution completes or if an error occurs.
+3. **DuckDB connection management** — Opens all 12 shard connections inside a `contextlib.ExitStack`, ensuring every connection is properly closed when resolution completes or if an error occurs.
 
 4. **Entity resolution** — Delegates to `fullmap.resolve()` which queries the sharded DuckDB database, ranks matches by category priority, preferred-name exactness, NLP level, and category frequency, then deduplicates to one CURIE per input string.
 
@@ -225,7 +225,7 @@ Both levels are queried during resolution. Level one (exact case-insensitive mat
 
 - If the `datassert` path does not contain the expected shard files, `duckdb.connect()` will raise an `IOException`.
 - If `entities` is empty, the function returns a dictionary with empty lists for all output columns.
-- The `ExitStack` ensures all 16 DuckDB connections are closed even if resolution raises an exception.
+- The `ExitStack` ensures all 12 DuckDB connections are closed even if resolution raises an exception.
 - Unresolved entities are silently filtered from the output (logged at INFO level by default via `resolve()`).
 
 ## Integration

{tablassert-7.3.1 → tablassert-7.3.3}/docs/configuration/graph.md

@@ -60,12 +60,14 @@ See [Table Configuration](table.md) for details.
 
 **`datassert: path`**
 
-Path to the datassert directory for entity resolution. Tablassert opens 16 shard files from `datassert/data/{0..15}.duckdb`. This database contains:
+Path to the [datassert](../datassert.md) directory for entity resolution. Tablassert opens 12 shard files from `datassert/data/{0..11}.duckdb`. This database contains:
 - Synonym mappings (text → CURIE)
 - Biolink categories
 - Taxonomic information
 - Source provenance (which database provided the mapping)
 
+See [Datassert](../datassert.md) for installation, build commands, and database schema.
+
 **`pubmed_db: path`**
 
 Optional path to SQLite database with PubMed metadata:
@@ -165,4 +167,5 @@ This processes a single table configuration (ALAMV6.yaml) into a knowledge graph
 ## Next Steps
 
 - **[Table Configuration](table.md)** - Learn how to define table transformations
+- **[Datassert](../datassert.md)** - Entity-resolution database installation and build
 - **[Tutorial](../tutorial.md)** - Complete example walkthrough

tablassert-7.3.3/docs/datassert.md

@@ -0,0 +1,106 @@
+# Datassert
+
+Datassert is a high-performance CLI for building a DuckDB-backed assertion store from NCATS Translator BABEL export files, with a focus on fast local builds and simple command-driven workflows. It produces the entity-resolution database used by Tablassert, containing biological synonyms, CURIEs, Biolink categories, taxon IDs, and source provenance, enabling `resolve()` to map free-text strings to standardized identifiers.
+
+## Installation
+
+```bash
+# Install CLI from GitHub
+go install github.com/SkyeAv/datassert@latest
+
+# Verify install
+datassert --help
+```
+
+## Build Command
+
+```bash
+# Build a Datassert database (downloads BABEL data automatically)
+datassert build
+```
+
+The build command automatically downloads BABEL exports from RENCI (`https://stars.renci.org/var/babel_outputs`), processes them, and produces sharded DuckDB databases.
+
+### Flags
+
+| Flag | Required | Default | Description |
+|------|----------|---------|-------------|
+| `--skip-downloads` / `-s` | No | `false` | Skip the BABEL download phase (use previously downloaded files) |
+| `--use-existing-parquets` / `-p` | No | `false` | Use existing Parquet files to rebuild DuckDB databases |
+
+### Data Pipeline
+
+1. **Download** — BABEL class and synonym files are downloaded from RENCI and split into LZ4-compressed NDJSON chunks under `./datassert/downloads/`.
+2. **Lookup** — Class files (`*.ndjson.lz4`) are read to build an in-memory equivalent-identifier lookup.
+3. **Parquet Staging** — Synonym files are processed with the lookup, quality-controlled, and written as sharded Parquet files to `./datassert/parquets/`.
+4. **DuckDB Generation** — Parquet files are loaded into 12 sharded DuckDB databases under `./datassert/data/`.
+
+### Examples
+
+```bash
+# Full build (download, process, and generate databases)
+datassert build
+
+# Skip downloads if BABEL files were already fetched
+datassert build --skip-downloads
+
+# Rebuild DuckDB databases from existing Parquet files
+datassert build --use-existing-parquets
+```
+
+### Runtime Behavior
+
+- Displays progress bars for download, class lookup, synonym processing, and DuckDB build phases.
+- Uses 90% of available CPUs for concurrent processing.
+- Downloads are retried up to 3 times on failure with a 10-second backoff.
+- All working files are stored under `./datassert/`.
+
+## Output Artifacts
+
+- 12 sharded DuckDB databases are written to `./datassert/data/{0..11}.duckdb`.
+- Each shard contains `SOURCES`, `CATEGORIES`, `CURIES`, and `SYNONYMS` tables, deduplicated, sorted, and indexed for query performance.
+- Staging Parquet files are written to `./datassert/parquets/{0..11}/`.
+
+Terms are routed to shards deterministically via `xxhash64(term) % 12`, so a given string always hits the same shard.
+
+### Schema
+
+Each shard contains four tables:
+
+| Table | Key Columns | Description |
+|-------|-------------|-------------|
+| `SYNONYMS` | `SYNONYM`, `CURIE_ID`, `SOURCE_ID` | Text synonym → CURIE mapping |
+| `CURIES` | `CURIE_ID`, `CURIE`, `PREFERRED_NAME`, `TAXON_ID`, `CATEGORY_ID` | Canonical identifiers and preferred names |
+| `CATEGORIES` | `CATEGORY_ID`, `CATEGORY_NAME` | Biolink category names |
+| `SOURCES` | `SOURCE_ID`, `SOURCE_NAME`, `SOURCE_VERSION` | Source database and version provenance |
+
+## Usage in Graph Config
+
+The `datassert:` field in a GC2 graph configuration points to the directory containing the shards. Tablassert opens all 12 shards at startup and passes the connections to `resolve()`.
+
+```yaml
+# graph-config.yaml (GC2)
+syntax: GC2
+name: my-graph
+version: "1.0"
+datassert: /path/to/datassert/   # directory containing data/0..11.duckdb
+tables:
+  - ./TABLE/my-table.yaml
+```
+
+## Programmatic Usage
+
+When calling `resolve()` directly, open the shard connections yourself:
+
+```python
+import duckdb
+from tablassert.fullmap import resolve
+
+datassert_dir = "/path/to/datassert"
+conns = [
+    duckdb.connect(f"{datassert_dir}/data/{i}.duckdb", read_only=True)
+    for i in range(12)
+]
+```
+
+See [Entity Resolution](api/fullmap.md) for the full `resolve()` API.

{tablassert-7.3.1 → tablassert-7.3.3}/docs/docker.md

@@ -81,8 +81,8 @@ docker run --rm \
 
 - **Datassert path** — The graph configuration YAML specifies the `datassert` path for the entity-resolution database. Ensure it is accessible inside the container.
 - **Multiprocessing** — `src/tablassert/cli.py:63` uses `multiprocessing.Pool` for parallel table loading and section extraction.
-- **DuckDB connections** — An `ExitStack` at `src/tablassert/cli.py:81` opens read-only connections to all 16 Datassert DuckDB shards concurrently.
-- **Entity resolution** — The `fullmap` module (`src/tablassert/fullmap.py`) shards terms across 16 DuckDB shards (`SHARDS = 16`) using xxhash64.
+- **DuckDB connections** — An `ExitStack` at `src/tablassert/cli.py:81` opens read-only connections to all 12 Datassert DuckDB shards concurrently.
+- **Entity resolution** — The `fullmap` module (`src/tablassert/fullmap.py`) shards terms across 12 DuckDB shards (`SHARDS = 12`) using xxhash64.
 - **Text normalization** — `src/tablassert/nlp.py` provides `level_one` (strip + lowercase) and `level_two` (regex-based cleanup).
 
 ## CI/CD Integration

{tablassert-7.3.1 → tablassert-7.3.3}/llms.txt

@@ -4,7 +4,7 @@
 This file is for two audiences: (1) YAML configuration authors and (2) package contributors.
 When source code and prose docs disagree, treat `src/tablassert/models.py` and `src/tablassert/cli.py` as the current authority.
 If you encounter older configurations, migrate `dbssert` to `datassert` (directory path).
-Current CLI behavior opens shard files at `datassert/data/{0..15}.duckdb`.
+Current CLI behavior opens shard files at `datassert/data/{0..11}.duckdb`.
 
 ## Quickstart
 - [README](README.md): high-level overview, install snippets, and one-command graph build.

{tablassert-7.3.1 → tablassert-7.3.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "tablassert"
-version = "7.3.1"
+version = "7.3.3"
 description = "Extract knowledge assertions from tabular data into NCATS Translator-compliant KGX NDJSON — declaratively, with entity resolution and quality control built in."
 authors = [
     { name = "Skye Lane Goetz", email = "sgoetz@isbscience.org" }
@@ -56,7 +56,7 @@ dependencies = [
     "scikit-learn>=1.8.0",
     "sentence-transformers>=5.3.0",
     "sqlite-utils>=3.39",
-    "typer>=0.24.1",
+    "typer>=0.21.2",
     "xxhash>=3.6.0",
 ]
 

{tablassert-7.3.1 → tablassert-7.3.3}/src/tablassert/fullmap.py

@@ -16,7 +16,7 @@ else:
     plh = Lazy.load("polars_hash")
 
 
-SHARDS: int = 16
+SHARDS: int = 10
 
 
 def empty_matches(column_context: bool) -> pl.DataFrame:
@@ -39,15 +39,15 @@ def empty_matches(column_context: bool) -> pl.DataFrame:
     return pl.DataFrame(schema=schema)  # pyright: ignore
 
 
-def distinct(lf: pl.LazyFrame, l0: str, l1: str, col: str = "term") -> pl.LazyFrame:
+def distinct(lf: pl.LazyFrame, l1: str, l2: str, col: str = "term") -> pl.LazyFrame:
     # ? Extract Unique Terms From Two Text Normalization Columns As LazyFrame
-    t0: pl.LazyFrame = lf.select(pl.col(l0).alias(col)).unique()
-    t0 = t0.with_columns(pl.lit(0).alias("nlp level"))
-
     t1: pl.LazyFrame = lf.select(pl.col(l1).alias(col)).unique()
     t1 = t1.with_columns(pl.lit(1).alias("nlp level"))
 
-    terms: pl.LazyFrame = pl.concat([t0, t1]).unique(subset=[col], keep="first")
+    t2: pl.LazyFrame = lf.select(pl.col(l2).alias(col)).unique()
+    t2 = t2.with_columns(pl.lit(2).alias("nlp level"))
+
+    terms: pl.LazyFrame = pl.concat([t1, t2]).unique(subset=[col], keep="first")
 
     bad: str = r"^\d+$|^(none|nan|na|null|unknown)$|^$"
     terms = terms.filter(~pl.col(col).str.contains(bad))
@@ -172,10 +172,10 @@ def resolve(
     tag: str = " two",
 ) -> pl.LazyFrame:
     # ? Case Dependant, Provenance Rich Name Entity Recognition
-    l0: str = col
-    l1: str = add(l0, tag)
+    l1: str = col
+    l2: str = add(l1, tag)
 
-    terms: pl.LazyFrame = distinct(lf, l0, l1)
+    terms: pl.LazyFrame = distinct(lf, l1, l2)
     matches: pl.DataFrame = query_distinct(terms, conns, taxon, prioritize, avoid, column_context)
 
     if log:
@@ -184,45 +184,45 @@ def resolve(
     # ! Collection Point: Join After DuckDB Query, Then Re-Lazy
     df: pl.DataFrame = lf.collect()
     result: pl.DataFrame = df.join(
-        matches.filter(pl.col("NLP_LEVEL").eq(0)), left_on=l0, right_on="term", how="left", suffix=" l0"
+        matches.filter(pl.col("NLP_LEVEL").eq(1)), left_on=l1, right_on="term", how="left", suffix=" l1"
     )
 
-    l1_matches: pl.DataFrame = matches.filter(pl.col("NLP_LEVEL").eq(1))
-    result = result.join(l1_matches, left_on=l1, right_on="term", how="left", suffix=" l1")
+    l2_matches: pl.DataFrame = matches.filter(pl.col("NLP_LEVEL").eq(2))
+    result = result.join(l2_matches, left_on=l2, right_on="term", how="left", suffix=" l2")
 
     result = result.with_columns(
         [
-            pl.when(pl.col("CURIE").is_not_null()).then(pl.col("CURIE")).otherwise(pl.col("CURIE l1")).alias(col),
+            pl.when(pl.col("CURIE").is_not_null()).then(pl.col("CURIE")).otherwise(pl.col("CURIE l2")).alias(col),
             pl.when(pl.col("PREFERRED_NAME").is_not_null())
             .then(pl.col("PREFERRED_NAME"))
-            .otherwise(pl.col("PREFERRED_NAME l1"))
+            .otherwise(pl.col("PREFERRED_NAME l2"))
             .alias(add(col, " name")),
             pl.when(pl.col("CATEGORY_NAME").is_not_null())
             .then(add(pl.lit("biolink:"), pl.col("CATEGORY_NAME")))
-            .otherwise(add(pl.lit("biolink:"), pl.col("CATEGORY_NAME l1")))
+            .otherwise(add(pl.lit("biolink:"), pl.col("CATEGORY_NAME l2")))
             .alias(add(col, " category")),
             pl.when(pl.col("TAXON_ID").is_not_null())
             .then(add(pl.lit("NCBITaxon:"), pl.col("TAXON_ID").cast(pl.String)))
-            .otherwise(add(pl.lit("NCBITaxon:"), pl.col("TAXON_ID l1").cast(pl.String)))
+            .otherwise(add(pl.lit("NCBITaxon:"), pl.col("TAXON_ID l2").cast(pl.String)))
             .alias(add(col, " taxon")),
             pl.when(pl.col("SOURCE_NAME").is_not_null())
             .then(pl.col("SOURCE_NAME"))
-            .otherwise(pl.col("SOURCE_NAME l1"))
+            .otherwise(pl.col("SOURCE_NAME l2"))
             .alias(add(col, " source")),
             pl.when(pl.col("SOURCE_VERSION").is_not_null())
             .then(pl.col("SOURCE_VERSION"))
-            .otherwise(pl.col("SOURCE_VERSION l1"))
+            .otherwise(pl.col("SOURCE_VERSION l2"))
             .alias(add(col, " source version")),
             pl.when(pl.col("NLP_LEVEL").is_not_null())
             .then(pl.col("NLP_LEVEL"))
-            .otherwise(pl.col("NLP_LEVEL l1"))
+            .otherwise(pl.col("NLP_LEVEL l2"))
             .alias(add(col, " nlp level")),
         ]
     )
 
     result = result.select(
         pl.exclude(
-            r"^(CURIE|PREFERRED_NAME|CATEGORY_NAME|TAXON_ID|SOURCE_NAME|SOURCE_VERSION|NLP_LEVEL|PR|FREQUENCY)( l1)?$"
+            r"^(CURIE|PREFERRED_NAME|CATEGORY_NAME|TAXON_ID|SOURCE_NAME|SOURCE_VERSION|NLP_LEVEL|PR|FREQUENCY)( l2)?$"
         )
     )
     result = result.select(pl.exclude(add(col, " two")))

{tablassert-7.3.1 → tablassert-7.3.3}/uv.lock

@@ -2211,7 +2211,7 @@ wheels = [
 
 [[package]]
 name = "tablassert"
-version = "7.3.0"
+version = "7.3.3"
 source = { editable = "." }
 dependencies = [
     { name = "duckdb" },
@@ -2276,7 +2276,7 @@ requires-dist = [
     { name = "sentence-transformers", specifier = ">=5.3.0" },
     { name = "sqlite-utils", specifier = ">=3.39" },
     { name = "tablassert", extras = ["rtcompat"], marker = "extra == 'rt'" },
-    { name = "typer", specifier = ">=0.24.1" },
+    { name = "typer", specifier = ">=0.21.2" },
     { name = "xxhash", specifier = ">=3.6.0" },
 ]
 provides-extras = ["rtcompat", "rt"]

tablassert-7.3.1/docs/datassert.md

@@ -1,66 +0,0 @@
-# Datassert
-
-Datassert is the entity-resolution database used by Tablassert. It contains biological synonyms, CURIEs, Biolink categories, taxon IDs, and source provenance, enabling `resolve()` to map free-text strings to standardized identifiers.
-
-## Installation
-
-```bash
-git clone https://github.com/SkyeAv/datassert
-```
-
-## Structure
-
-Datassert is split into 16 DuckDB shard files for parallel querying:
-
-```
-datassert/
-  data/
-    0.duckdb
-    1.duckdb
-    ...
-    15.duckdb
-```
-
-Terms are routed to shards deterministically via `xxhash64(term) % 16`, so a given string always hits the same shard.
-
-### Schema
-
-Each shard contains four tables:
-
-| Table | Key Columns | Description |
-|-------|-------------|-------------|
-| `SYNONYMS` | `SYNONYM`, `CURIE_ID`, `SOURCE_ID` | Text synonym → CURIE mapping |
-| `CURIES` | `CURIE_ID`, `CURIE`, `PREFERRED_NAME`, `TAXON_ID`, `CATEGORY_ID` | Canonical identifiers and preferred names |
-| `CATEGORIES` | `CATEGORY_ID`, `CATEGORY_NAME` | Biolink category names |
-| `SOURCES` | `SOURCE_ID`, `SOURCE_NAME`, `SOURCE_VERSION` | Source database and version provenance |
-
-## Usage in Graph Config
-
-The `datassert:` field in a GC2 graph configuration points to the directory containing the shards. Tablassert opens all 16 shards at startup and passes the connections to `resolve()`.
-
-```yaml
-# graph-config.yaml (GC2)
-syntax: GC2
-name: my-graph
-version: "1.0"
-datassert: /path/to/datassert/   # directory containing data/0..15.duckdb
-tables:
-  - ./TABLE/my-table.yaml
-```
-
-## Programmatic Usage
-
-When calling `resolve()` directly, open the shard connections yourself:
-
-```python
-import duckdb
-from tablassert.fullmap import resolve
-
-datassert_dir = "/path/to/datassert"
-conns = [
-    duckdb.connect(f"{datassert_dir}/data/{i}.duckdb", read_only=True)
-    for i in range(16)
-]
-```
-
-See [Entity Resolution](api/fullmap.md) for the full `resolve()` API.